fix(redis): ask INFO, not CLUSTER INFO, whether this is a cluster

[giokur]

cluster_enabled lives in INFO's Cluster section. CLUSTER INFO reports the state of a
cluster — cluster_state, cluster_slots_assigned, cluster_known_nodes — and never carries
cluster_enabled. So the probe added in #2348

jedis.clusterInfo().contains("cluster_enabled:1")   // false against every server alive

always answered "not a cluster".

Why it got through

The wrong answer is the right answer on a single node. dev and qa are single-node Memorystore, so the
probe looked correct there and the password-reset lookup kept working.

Production is a six-node cluster. There, clusterMode() returned false, getResetToken took the
standalone branch, SCAN ran against one seed and the first MGET came back:

JedisMovedDataException: MOVED 4478 redis-cluster-4.internal.openframe.ai:6379
    at redis.clients.jedis.UnifiedJedis.mget(UnifiedJedis.java:912)
    at com.openframe.test.data.redis.Redis.findToken(Redis.java:100)
    at com.openframe.test.data.redis.Redis.getResetToken(Redis.java:82)   ← standalone branch

46 times across the rotator's 90-second poll, which surfaces as No password-reset token for test@flamingo.cx appeared within 90s (46 scans). The production tenant report failed its 16:00 run on
2026-09-23 — the first after 6.36.10 rolled out — and every later run would have failed the same way.
Production keeps no fallback password for that account by design, so a failed rotation stops the report
entirely rather than degrading it.

The prod log line that names the bug outright:

Redis at redis-cluster-0.redis-cluster-headless.datasources.svc.cluster.local:6379
  reports cluster mode disabled

against a server that answers MOVED. That is the probe calling a live cluster a single node.

The fix

jedis.info("cluster"), whose Cluster section is where cluster_enabled actually is.

Everything else is unchanged: a probe that cannot connect still assumes a cluster for that one lookup
and remembers nothing, and RedisConfig.getConfiguredCluster() still wins.

Compiles.

Ships with saas-shared, which pins test.redis.cluster: true in configs/prod so production is
restored at its next config refresh rather than waiting on this release, the pin bump and a rollout.
That pin can come out once this is deployed.

---

🤖 Generated with Claude Code

[github-actions]

🦩 Flamingo Code Review

No findings on the current head.

Mode: advisory

---

Need another pass? Commits pushed after this review are not reviewed automatically.

• Review the new commits — the commits added since this review
• Review the whole diff again — ignoring what was already reviewed

Prefer typing? Comment @flamingo-review, or @flamingo-review full. To review every push on this pull request, add the flamingo-review-always label.

React 👍/👎 on inline comments to teach the reviewer.

Started 2026-09-23 16:41 UTC · updated 2026-09-23 16:41 UTC · workflow run